[SOLR-13044] IndexFetcher.openNewSearcherAndUpdateCommitPoint can throw NPE on null SolrCore

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 9.1
• Assignee: David Smiley
• Reporter: Chris M. Hostetter
• Created: 06/Dec/18 00:06
• Updated: 10/Jan/23 20:45
• Resolved: 27/Apr/22 16:30

Description

While working through the logs in SOLR-13031, i observed the following...

  [beaster]   2> 57230 ERROR (indexFetcher-193-thread-1) [    ] o.a.s.h.ReplicationHandler Index fetch failed :org.apache.solr.common.SolrException: Index fetch failed : 
  [beaster]   2>        at org.apache.solr.handler.IndexFetcher.fetchLatestIndex(IndexFetcher.java:676)
  [beaster]   2>        at org.apache.solr.handler.IndexFetcher.fetchLatestIndex(IndexFetcher.java:346)
  [beaster]   2>        at org.apache.solr.handler.ReplicationHandler.doFetch(ReplicationHandler.java:424)
  [beaster]   2>        at org.apache.solr.handler.ReplicationHandler.lambda$setupPolling$13(ReplicationHandler.java:1184)
  [beaster]   2>        at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
  [beaster]   2>        at java.util.concurrent.FutureTask.runAndReset(FutureTask.java:308)
  [beaster]   2>        at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:180)
  [beaster]   2>        at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:294)
  [beaster]   2>        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
  [beaster]   2>        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
  [beaster]   2>        at java.lang.Thread.run(Thread.java:748)
  [beaster]   2> Caused by: java.lang.NullPointerException
  [beaster]   2>        at org.apache.solr.handler.IndexFetcher.openNewSearcherAndUpdateCommitPoint(IndexFetcher.java:909)
  [beaster]   2>        at org.apache.solr.handler.IndexFetcher.fetchLatestIndex(IndexFetcher.java:653)
  [beaster]   2>        ... 10 more

Which comes from this bit of code...

    // must get the latest solrCore object because the one we have might be closed because of a reload
    // todo stop keeping solrCore around
    SolrCore core = solrCore.getCoreContainer().getCore(solrCore.getName());
    try {
      Future[] waitSearcher = new Future[1];
      searcher = core.getSearcher(true, true, waitSearcher, true);
      if (waitSearcher[0] != null) {
        try {
          waitSearcher[0].get();
        } catch (InterruptedException | ExecutionException e) {
          SolrException.log(log, e);
        }
      }
      commitPoint = searcher.get().getIndexReader().getIndexCommit();
    } finally {
      if (searcher != null) {
        searcher.decref();
      }
      core.close(); // IndexFetcher.java:909 where NPE is being thrown
    }

IIUC: openNewSearcherAndUpdateCommitPoint isn't accounting for the possibility that the the IndexFetcher is getting run before the SolrCore has been fully initialized and regsitere with the CoreContainer and/or after/while the SolrCore (and/or CoreContainer) is in the process of being closed, and isn't available from solrCore.getCoreContainer().getCore(String)

Issue Links

• relates to: SOLR-13031 Harden ChaosMonkeySafeLeaderWithPullReplicasTest
• links to: GitHub Pull Request #794
• links to: GitHub Pull Request #815